Utilization climate hazards group infrared precipitation with station data (CHIRPS) to determine rainfall thresholds for early warning landslide in Luwu, South Sulawesi, Indonesia

Abstract

Early warning of landslides using remote sensing has received increasing attention in recent decades. However, in Luwu Raya, South Sulawesi, rainfall thresholds that trigger landslides remain poorly understood due to limited rainfall ground data. This study evaluates the suitability of the Climate Hazards Group InfraRed Precipitation with Station (CHIRPS) to determine landslide-triggering thresholds. This lack of understanding poses significant risks to local communities and infrastructure, making it crucial to identify reliable rainfall thresholds for landslide prediction and risk mitigation. Because satellite rainfall estimates have varying biases, both spatially and temporally, validation was performed using observational data from 41 ground-based rainfall observation as reference data for 2019–2024 against CHIRPS. Accuracy was assessed using RMSE, a dichotomous metric, and extreme rainfall indices. Rainfall thresholds were derived by event-based matching between the landslide inventory and CHIRPS rainfall, followed by curve analysis of daily rainfall combined with antecedent accumulations over 3, 5, 10, 15, 20, and 30 days. Thresholds were derived from curve analysis of accumulated rainfall (3–30 days) based on flood and landslide events. Results show that CHIRPS accuracy is lower in mountainous areas (>1000 m) than in lowlands. The most reliable threshold combined daily rainfall with 15-day antecedent rainfall, achieving 67% accuracy, while shorter accumulation (10 days) performed poorly (5%). Central Luwu (500–750 m) showed the highest predictive potential. The identified threshold of only 20 mm/day, categorized as light rainfall in Indonesia, suggests that while CHIRPS offers promise, further refinement is needed before satellite rainfall can be fully applied in operational landslide early warning.